Method of embedding digital watermark, storage medium in which the method is stored, method of identifying embedded digital watermark, and apparatus for embedding digital watermark

Abstract

The procedure of the present invention specifies predetermined elements included in a matrix F which is obtained by Fourier transform of master image data P0 (step S122), and adds a minute variation ΔF of a predetermined magnitude to either a real number array FR or an imaginary number array FI of the predetermined elements (step S124 and S126). Here note that the minute variation ΔF should be subtracted from corresponding elements, in order to keep the symmetry of the Fourier spectrum. A resulting image obtained by an inverse transform of the matrix with the minute variation ΔF added thereto includes a phase difference pattern W01 that is embedded therein and corresponds to the minute variation ΔF. As long as a master image is closed to the public, this embedded pattern can not be taken out of the resulting image nor be deleted by overwriting attacks. Even if the processed data with the digital watermark embedded therein is exposed to overwriting attacks of different pieces of watermark information by the similar algorithm, the arrangement of the information enables the digital watermark embedded in the master data to be taken out accurately. The similar series of processing may be carried out with regard to a specific area of the low frequency component obtained by wavelet transform of the mater image.

Description

TECHNICAL FIELD[0001]The present invention relates to a digital watermarking technique and more specifically to a technique that embeds watermark information in data, such as image data.BACKGROUND ART[0002]A variety of digital watermarking techniques have been proposed to embed copyright information in digitalized copies of images and music. Digitalized data is readily copied in a complete form (namely, digitalized data has perfect reproducibility), so that the protecting measure is required against illegal copies. The digital watermarking technique electronically embeds watermark information, such as copyright information, in master data in a human-imperceptible manner. The embedded watermark information can be taken out according to the requirements. The digital watermark clearly shows any third party the presence of the copyright in digitalized copies of images and music. The digital watermark generally includes information for identifying the copyright holder. The data with the ...

AI Technical Summary

Benefits of technology

[0005]The object of the present invention is to solve the drawbacks of the prior art technique discussed above and thus to provide an embedding technique that ensures the good preservation of a legal digital watermark and the high secrecy of the digital watermark to any third party even when electronic data open to the public is exposed to overwriting attacks and data compression.

[0008]This technique utilizes the characteristics that a variation of either a real number array or an imaginary number array in a spectrum, which is obtained by discrete Fourier transform of master data in a frequency domain, leads to a variation in phase. The technique of the present invention adds a minute variation corresponding to a phase difference pattern to either the real number array or the imaginary number array obtained by the discrete Fourier transform of the master data and subsequently carriers out an inverse transform of the processed data, so as to generate authorized data with watermark information embedded therein. The phase difference pattern is taken out through comparison between the master image data and the authorized data. This arrangement enables the user to intuitively grasp the embedded digital watermark as the phase difference pattern. As long as the master data without the digital watermark embedded therein is not identified, the phase difference pattern can be taken out accurately even after overwriting attacks.

[0010]Here it is favorable to carry out the addition of the minute variation ΔF to either the real number array or the imaginary number array while keeping the symmetry of either the real number array or the imaginary number array. As the results of the discrete Fourier transform, the real number array has even symmetry and the imaginary number array has odd symmetry. The addition of the minute variation by considering these points enables good preservation of the intrinsic characteristics of the master data.

[0012]It is also favorable to add the minute variation to a component in a low frequency domain of the real number array or the imaginary number array. In the case of addition to the high frequency component, data compression like JPEG may result in losing the watermark information. In the case of addition to the low frequency domain, on the other hand, data compressing does not cause the phase difference pattern as the digital watermark to be lost. The phase difference pattern may, however, be added to the high frequency domain in the case where consideration is given to only reversible compression methods that store the data in the course of compression and restore the original information to a perfect form by expansion of the data.

[0015]In order to add the minute variation in such a manner that the digital watermark is not lost by data compression, one applicable method carries out a predetermined data transform to obtain a specific data portion mainly corresponding to a low frequency component from master data, prior to discrete Fourier transform of master data. The method then causes the specific data portion to be subjected to discrete Fourier transform and adds a minute variation corresponding to a predetermined phase difference pattern as watermark information to either one of a real number array or an imaginary number array obtained by the discrete Fourier transform. After the addition of the minute variation, the method carriers out inverse Fourier transform and subsequently an inverse transform of the predetermined data transform. The combination of the data compression with the discrete Fourier transform enables the digital watermark to be accurately embedded in the low frequency domain. This method ensures the high resistance against the overwriting attacks, which is discussed above as the effect of the method of embedding the digital watermark by the discrete Fourier transform, and also effectively prevents the digital watermark data from being altered or deleted by the high degree of data compression. A typical example of the data transform and the inverse data transform is wavelet transform and inverse wavelet transform. Using the wavelet transform and the inverse wavelet transform desirably simplifies the procedures of embedding and restoring watermark information. Any of various known techniques is applicable for the wavelet transform here. A typical example is orthogonal wavelet transform using the Haar basis. Any data transform and inverse data transform other than the wavelet transform and the inverse wavelet transform is also applicable for this method, as long as it follows a transform algorithm to obtain a specific data portion mainly corresponding to the low frequency domain.

[0026]Even when the plurality of overwriting attacks are made, these methods enable extraction of the phase difference pattern legally embedded in the master data and easy identification of the legal data.

Problems solved by technology

The prior art digital watermarking technique, however, has the drawback that the legal watermark information is not taken out accurately when another piece of information overwrites the existing information by a similar procedure.

When the authorized data is exposed to the overwriting attack to prevent the embedded legal digital watermark from being normally read, the digital watermarking loses its significance.

The digital watermark data embedded by the prior art digital watermarking technique, however, may be altered or even deleted by such data compression.

Namely the prior art technique is not sufficiently practical.

Images